The puzzle here is that better safety training was expected to bring accidents down, but the survey shows accidents going up right after training improved. Resolving this means finding a fact that removes the contradiction rather than one that just adds detail.
Only the third option removes the assumption that training came first and accidents rose afterward because of it. Once we know manufacturers only upgrade training after accidents are already rising, the numbers in the survey make sense without any conflict with what experts expect.
Let's summarize:
So the discrepancy is best explained by option 3, manufacturers improve training only after accidents have already started rising.